Author + information
- Peter Huckfeldt, PhD,
- José Escarce, MD, PhD,
- Andrew Wilcock, PhD,
- Neeraj Sood, PhD,
- Brendan Rabideau, BA,
- Ioana Popescu, MD, MS and
- Teryl Nuckols, MD, MSHS∗ (, )@terylnuckols@CedarsSinai
- ↵∗Cedars-Sinai Medical Center, 8700 Beverly Boulevard, Becker 113, Los Angeles, California 90048
After announcement and implementation of the Medicare Hospital Readmissions Reduction Program (HRRP), 30-day readmissions declined rapidly among seniors with heart failure (HF) while 30-day mortality rose (1). This raised questions about whether the policy is responsible, because lower HF readmission rates have historically been associated with higher mortality (2). Because penalized hospitals faced more pressure to reduce readmissions and exhibited larger declines, we compared trends in HF mortality at penalized and nonpenalized hospitals nationally (3).
We performed interrupted time-series analyses of risk-adjusted 30-day all-cause mortality and unplanned readmission rates during “pre-HRRP” (January 2007 to March 2010), “anticipation” (February 2010 to September 2012), and “penalty” (October 2012 to November 2014) periods. Subjects included Medicare beneficiaries age ≥65 years with index discharges for HF, with HRRP exclusions. Study hospitals were eligible for the HRRP in fiscal year (FY) 2013 or FY2014. Data sources included 2007 to 2014 Medicare Provider and Analysis Review, Master Beneficiary Summary, Provider of Services, and HRRP payment adjustment datasets (4). The unit of analysis was an index discharge. We used regression models to estimate changes in outcomes under the HRRP relative to a linear pre-HRRP time trend, allowing changes in outcomes and pre-HRRP trends to differ by penalty status. Models controlled for patient characteristics, season, and hospital fixed effects, and accounted for clustering within hospitals.
To test whether outcomes changed under the HRRP, we used the regression estimates to compare observed and projected outcomes (Δ) at penalized and nonpenalized hospitals during the penalty period. To assess whether changes differed between penalized and nonpenalized hospitals, we compared their observed-to-projected differences (ΔΔ). These comparisons are a robust way of testing whether the slope and intercept of time trends changed at any point after the pre-HRRP period.
Among 3,256 eligible hospitals, 2,214 (68.00%) incurred penalties in FY2013. Of 3,009,296 index discharges across the study periods, 2,248,665 (74.72%) occurred at penalized hospitals. Figure 1 displays regression estimates graphically, plotting observed risk-adjusted outcomes along with projections of what outcomes would have been during anticipation and penalty periods without the HRRP.
Pre-HRRP rates were 8.08% at penalized hospitals and 8.76% at nonpenalized hospitals. After announcement of the HRRP, observed mortality rates rose relative to pre-HRRP trends, exceeding projections by 0.62 percentage points (Δ95% confidence interval [CI]: 0.30 to 0.94; p < 0.001) at penalized hospitals and 0.60 points (Δ95% CI: 0.01 to 1.19; p = 0.047) at nonpenalized hospitals by the penalty period. Increases were similar at penalized and nonpenalized hospitals (ΔΔ 0.02 percentage points; 95% CI: −0.65 to 0.70; p = 0.949).
Pre-HRRP rates were 24.26% and 20.07% at penalized and nonpenalized hospitals, respectively. After announcement of the HRRP, observed readmission rates declined relative to pre-HRRP trends at penalized hospitals, reaching 3.35 percentage points (Δ95% CI: 3.90 to 2.81; p < 0.001) below projections by the penalty period. At nonpenalized hospitals, rates decreased during the pre-HRRP period and stabilized during the anticipation period, such that by the penalty period, readmissions exceeded projections by 1.27 percentage points (Δ95% CI: 0.44 to 2.10; p = 0.003). Declines relative to projections were 4.62 percentage points (ΔΔ95% CI: 5.62 to 3.63; p < 0.001) larger at penalized than nonpenalized hospitals.
The fact that HF readmissions declined at penalized hospitals but did not decline at nonpenalized hospitals suggests that they responded differently to the HRRP. By contrast, mortality rose similarly at penalized and nonpenalized hospitals. Effective responses to the HRRP, such as improving transition-related care, could reduce both readmissions and mortality. Ineffective responses may create barriers to inpatient care, decreasing readmissions but potentially increasing mortality.
It is tempting to attribute rising HF mortality to the HRRP. However, the similarity in mortality trends at penalized and nonpenalized hospitals does not support the hypothesis that the HRRP led to increased mortality, given the divergent trends in readmissions. In this context, any conclusion that the HRRP increased mortality at penalized hospitals is speculative.
A secular shift in the HF population, not fully accounted for by standard risk-adjustment techniques, could be contributing to parallel increases in mortality at penalized and nonpenalized hospitals. Age-adjusted mortality rates in patients with HF increased sharply from 2012 to 2014, irrespective of setting, and more of these deaths were from noncardiovascular causes (5). The causes of rising mortality among recently discharged patients with HF warrant further investigation.
Please note: This work is supported by the Agency for Healthcare Research and Quality (R01 HS024284) and National Institute on Aging (R01AG046838). The authors have reported that they have no relationships relevant to the contents of this paper to disclose. The authors thank Margaret Kelley, MSEd, Cedars-Sinai Medical Center; Jillian Wallis, PhD, University of Southern California; and Zhiyou Yang, BS, University of Minnesota, for their contributions to the research effort.
- 2018 American College of Cardiology Foundation
- Gupta A.,
- Allen L.A.,
- Bhatt D.L.,
- et al.
- Desai N.R.,
- Ross J.S.,
- Kwon J.Y.,
- et al.
- ↵Centers for Medicare & Medicaid Services, Hospital Readmissions Reduction Program. Available at: https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/AcuteInpatientPPS/Readmissions-Reduction-Program.html. Accessed September 28, 2017.
- Ni H.,
- Xu J.